Hi Bong,
Yup uptime requirements need to be defined, also I think its important to find
out:
1. Budget -- how much is the client willing to pay for HA (helps to know how
much do they lose if the site is down)
2. Rate of change -- high rate of change on the DB or content side can make HA
a challenge, if the rate of change is slow, rsync works :-)
3. for active/standy, what is the RTS -- what is the expected data loss/time
loss during failovers
4. application complexity -- some applications are inherently difficult to
replicate, kahit master-master
5. target user -- if you have external or internal users, which one is more
important to put in updates
Pero kung ang target mo is more a 'learning experience' -- implement mo muna
yung heartbeat, tapos subukan mo yung list ni Ariz :-)
For testing decide what you want to 'fail' -- network, disk, etc. tapos just
test your set up to see what works (or breaks).
Cheers,
JayJay
________________________________
From: fooler mail <[email protected]>
To: Caloocan Gangsta <[email protected]>; Philippine Linux Users' Group
(PLUG) Technical Discussion List <[email protected]>
Sent: Sunday, March 3, 2013 2:35 AM
Subject: Re: [plug] HIgh availability requirements
On Thu, Feb 28, 2013 at 11:13 PM, Caloocan Gangsta <[email protected]> wrote:
> Hi Pluggers,
>
> Would like to get your suggestion and howtos in providing high availability
> on my two server that will host Apache, i need to have a redundant
> connectivity even one of the server will fail.
>
> Thanks,
>
> Bong
hi bong,
how many 9's uptime percentage are you trying to achieve? The more 9s
you have the more redundant of network devices you need such us
redundant routers, load balancers, switches, servers, storage as well
as your upstream links...
for enterprise designs... not only those above are redundant.... even
the locations too and implementing global load balancing...
there are two types of highly availability - active-active and active-standby
active-active full use of your capacity while active-standby use one
half of your capacity.. highly availability is different with highly
scalability...
even though you have servers doing a highly availability but your
upstream link, router or load balancer is only one or not in highly
availability setup and one of them fails... your number of 9s will be
affected to this... that is why you need to understand your uptime
needs....
for server highly availability.. you go for hardware based load
balancers (eg. foundry, alteon, F5, cisco, etc) or software based load
balancers (eg. LVS)..
going for load balancer solution either hardware or software based
requires additional hardware investment..
my last design there in Philippine ISP before I left the country is to
implement a highly availability and scalability to all regions we have
point of presence (POP) without using a load balancer due to budget
constraint... services running in POP with highly availability and
scalability are DNS and PROXY both in active-active setup with 1
second failover to a good active server.. i wrote a customize C
network program to mimic how load balancer do a health check to a
service with a combination of one existing network protocol to help
the availability of layer 3 health status....
fooler.
_________________________________________________
Philippine Linux Users' Group (PLUG) Mailing List
http://lists.linux.org.ph/mailman/listinfo/plug
Searchable Archives: http://archives.free.net.ph
_________________________________________________
Philippine Linux Users' Group (PLUG) Mailing List
http://lists.linux.org.ph/mailman/listinfo/plug
Searchable Archives: http://archives.free.net.ph
